image shape conciliation

[luke14free]

I am re-implementing part of the model with vgg-16 and it's not clear to me how you are feeding the train dataset which is 800x600x3 into vgg-x which takes 224x224x3 images. I tried looking at your FCN.py code but it's still a mistery to me

[igorbarinov]

@luke14free I'd like to join your efforts in porting. Do you plan to work on it in open source way?

[luke14free]

@igorbarinov To be honest I am currently playing lego using Keras very little value on my side. I use the functions to retrieve the dataset from this project (BatchDatset, TestDataset classes), used the VGG16(weights='imagenet', include_top=False) from keras applications with a custom 800x600x3 Input and then adapted the code from https://github.com/JihongJu/keras-fcn to get the final part of the FCN-8 network. Now I am currently stuck on a stupid thing but as soon as it compiles I will open source it.

[dongwu92]

There are all conv layers and pool layers in vgg-x nets. And I simply force it to be applied in 800x600x3 images, and it may leads to the low accuracy of my code.